• Open Access

Bioenergy production and Skylark (Alauda arvensis) population abundance – a modelling approach for the analysis of land-use change impacts and conservation options

Authors


Abstract

Bioenergy production is seen as one way of meeting future energy needs. The growing demand for biomass for energy production induces the cultivation of a few fast growing and high-yielding energy crops on vast areas of arable land. This land-use change has been found associated with the reduction of habitat suitability for farmland birds and a decline in farmland biodiversity in general. A large number of studies have assessed the ecological effects of energy crop cultivation at the local scale of a single field. This study focuses on regional landscape changes caused by increased energy crop cultivation, which includes reduction of crop-type richness and spatial concentration of single crop-types. We present a spatially explicit ecological model to assess the population-level consequences of these effects on the abundance of the farmland bird species Skylark (Alauda arvensis). We also investigate the impacts of different land-use scenarios and aim to identify adaptive conservation options. We show that (1) the impacts of increased energy crop cultivation on Skylark population abundance depend strongly on the landscape structure; (2) impacts could be tolerated as long as a certain minimum level of crop-type heterogeneity is retained at the landscape level and (3) conservation actions are required and effective especially on landscapes where crop-field size is large.

Introduction

The production of renewable energy from energy crops is seen as an important option for mitigating climate change (Sims et al., 2006). This includes gaining independence from fossil fuel, reducing greenhouse gas emissions and sequestering carbon in soils. To reach these goals, policy makers in Europe, America and Asia encourage energy production from renewable sources. Targets of China and the European Union are an increase to 15% and 20% of primary energy by 2020, respectively, each including a large percentage of biomass (European Commission, 2008; Bringezu et al., 2009; for United States see the 2007 Energy Independence and Security Act EISA, 2007). However, bioenergy production is also controversial, as effects of large-scale energy crop production on biodiversity and the environment remain mostly unclear (Marland & Obersteiner, 2008; Dauber et al., 2010). Here, we focus on first-generation energy crops which are primarily grown for food (energy is made from sugar, starch and vegetable oil, Sims et al., 2008). Commercial second-generation biofuel production is not considered as it is in the early stages of application (Eisentraut, 2010; Fletcher et al., 2011).

The growing demand of renewable energy sources and political promotion of energy from biomass production (Bringezu et al., 2009) encourages the cultivation of a just few fast growing and high-yielding energy crops, e.g. canola, maize and green rye; the latter is rye harvested at heading, usually in May. These are cultivated on vast areas of arable land over relatively short timescales (USDA FAS, 2008; BMU, 2009; Fargione et al., 2010). Starting in 2007 after the repeal of a 1993 European Union regulation, energy crop cultivation on set-aside agricultural land in Europe particularly amplified this land-use change (Eurostat, 2009). Thus, first-generation bioenergy production from agricultural feedstock has triggered or amplified a large-scale land-use change, characterized mainly by (a) a low number of cultivated crop-types or rather the simplification of crop rotations at regional landscape scale resulting in large uniformly cropped areas and (b) a shift from spring- to autumn-sown cereals (EEA, 2007; Fargione et al., 2010; Lunetta et al., 2010; BMELV, 2011; Fletcher et al., 2011). Moreover, the proportion of energy crops has been found to be negatively associated with agricultural landscape diversity (Landis et al., 2008; Fletcher et al., 2011). To summarize, the strong pressure towards bioenergy crops has altered the composition of cultivated crops, the configuration of agricultural landscapes (heterogeneity and placement of fields) and local habitat suitability determined by crop characteristics (e.g. vegetation cover and height) as well as management practices, e.g. harvest techniques (Best, 1986; Dauber et al., 2010; Fletcher et al., 2011).

The loss of habitat diversity (heterogeneity) and the reduction of habitat suitability likely lead to a decline in farmland biodiversity (Benton et al., 2003; Eggers et al., 2009; Fargione et al., 2010; Fletcher et al., 2011). The decline in abundance of many farmland bird populations clearly illustrates this (Krebs et al., 1999; Newton, 2004; Sage et al., 2006). Bird population abundance generally responds rapidly to habitat changes in agricultural landscape and is often used as an indicator of biodiversity and environmental health (Gregory et al., 2004, 2005; Birrer et al., 2007; Dziewiaty & Bernardy, 2008; Davey et al., 2010).

A large number of studies have assessed the ecological effects of energy crop cultivation at the field scale (Dauber et al., 2010). However, examinations of bioenergy production impacts at the regional landscape scale are rare, and either mostly nonspatial (Butler et al., 2007; Landis et al., 2008; Eggers et al., 2009) or focused on a particular region with high effort at data collection (Haughton et al., 2009; Boatman et al., 2010; Zhang et al., 2010). For an ecological assessment of bioenergy landscape impacts, the spatial structure of production systems, including crop-type and related management, on regional scale likely matters as well (Hanowski et al., 1997; Zhang et al., 2010). Furthermore, regional bioenergy landscape types characterized by a differing average field size should be taken into account as it directly affects agricultural landscape diversity (e.g. in Germany average field size ranges from about 5 to 50 ha, Fischer et al., 2011).

This study aims to provide a better understanding of the effects of increasing energy crop cultivation on farmland birds at population-level. Wild bird indicators are well-developed proxies for the impacts of agricultural land-use change on biodiversity (Krebs et al., 1999; Gregory et al., 2005; British Trust for Ornithology, 2011a). Especially farmland birds are considered to be useful indicators because they occupy a wide range of habitats, tend to be near the top of the food chain and are responsive to farm-scale changes in habitat suitability (Ormerod & Watkinson, 2000; Henderson et al., 2009; British Trust for Ornithology, 2011a).

In order to assess the consequences of land-use change at landscape scale on Skylark (Alauda arvensis) population abundance, we developed and parameterized a fairly simple, spatially explicit ecological model (Chamberlain & Crick, 1999). The model evaluates the agricultural landscapes based on three main habitat requirements of the Skylark and produces two spatially explicit results, the habitat suitability for Skylark and the expected minimum number of Skylark breeding pairs. We analysed land-use change effects by (a) systematically reducing both the number of cultivated crop-types, which refers to a loss of habitat diversity, and the average crop-field sizes reflecting regional landscape types, and (b) simulating two land-use scenarios differing strongly in energy crop proportion. Based on both the call of nature conservation organizations for about 10% near-nature area in agricultural landscapes and the importance of structurally diverse landscapes, we aim to identify conservation options for Skylark populations (Vickery et al., 2009; Schoen, 2011).

Materials and methods

Skylark population abundance as indicator of biodiversity and wildlife health in the wider countryside

The model we present is parameterized for the Skylark (A. arvensis), which is a ground-breeding, central place foraging farmland bird species (British Trust for Ornithology, 2011b). We consider the Skylark to be a useful proxy for this kind of bird species as it (a) is well studied, (b) serves as indicator in studies elsewhere and (c) occupies a diversified set of habitat types related to farmland (UK wild bird indicator known as ‘Skylark index’ see Wilson et al., 1997; Gregory et al., 2004; Topping & Odderskaer, 2004; Boatman et al., 2010; British Trust for Ornithology, 2011b). Skylarks breed mainly from May to July with one to four broods and feed mostly on insects, but occasionally on cereal grain and weed seeds (Delius, 1965; Wakeham-Dawson et al., 1998; British Trust for Ornithology, 2011b).

Model structure

The modelling framework presented here is an ecological model that adopts elements from the habitat modelling approach (HM) commonly used in ecology to assess landscapes in terms of their impacts on the distribution and survival of a certain species (Schadt et al., 2002; Elith et al., 2006). Our model has a certain similarity to HM in the sense that the landscape is subdivided into a grid of patches, each containing information on relevant environmental factors (model input) and returning the calculated habitat suitability or bird abundance for the species of interest (model outcome). In our approach, however, the parameters characterizing crop-types summarize the cumulative effect of processes that occur over the course of one breeding season, influencing food availability and probability of breeding success. Major components of this modelling framework are a landscape generator and a protocol for habitat evaluation of a given landscape based on main requirements of the Skylark (A. arvensis) over one breeding season. The landscape generator creates several random landscape maps based on a few input parameters. The habitat evaluation protocol calculates both the spatial explicit habitat suitability and the species population abundance of each landscape map (see Fig. 2).

Landscape generator

The landscape generator presented here allowed us to systematically generate suites of virtual landscapes that coincide with typical spatial characteristics of agricultural landscapes (see Fig. 1). We defined an agricultural landscape as an irregular mosaic that consists of fields, mostly rectangular shaped, of a certain average size with a number of cultivated crop-types. Landscape details like roads, hedgerows and lakes were disregarded to focus on more general results of regional-scale agricultural land-use change effects (a knowledge gap discussed in Dauber et al., 2010). For the application of the model to a particular landscape, more landscape elements could easily be added.

Figure 1.

Real landscapes (left) and virtual landscapes as an example (right), each with an area of 3 × 3 km (Google, 2010a,b). Panels (a) and (b) depict high crop-type richness and small field size (average field size 3 ha and eight crop-types for virtual landscape). Panels (c) and (d) feature low crop-type richness and big field size (average field size of 43 ha and three crop-type for the virtual landscape).

The evaluation of land-use change focused on two objectives: the general impacts of both field size and crop-type richness using a random selection of crop-types, and two scenario-based landscapes with fixed values of both number of crop-types cultivated and crop-type area (details below). In this study, crop-type refers to (a) crop species and a certain production system and (b) Integrated Biodiversity Area (IBA) (see Table 1 for details). We defined IBA as near-nature area adapted to the regional environment without pesticide treatment and closely linked to extensive land use, hence with high plant and invertebrate diversity. This includes meadows, nonmanaged grassland habitats, open verges and flower strips without cutting operations between May and July (Siebert et al., 2010; Vepsäläinen et al., 2010; Cordeau et al., 2011; Deri et al., 2011). A crop-type subsumes both the characteristics of the crop species and the production regime that may affect habitat suitability of farmland birds during the breeding season (see section Crop-type parameter estimation, sensitivity analysis and model validation).

Table 1. Crop-types considered in this study and their habitat suitability characteristics for Skylark. A crop-type represents the characteristics of both a crop species and its production system. Landscape cover of crop-types in the Current scenario relates to the region of Prignitz in NE Germany (Dziewiaty & Bernardy, 2008); IBA, Integrated Biodiversity Area; SRC, short-rotation coppice willow plantation; rye, green rye I and green rye II differ in their production system, green rye is harvested at heading. For the BioenergyPlus scenario, we considered the creation of ‘bioenergy regions’ with a strong focus on regional energy supply from regional biomass sources (BioRegions, 2010)
Crop-typeProbability of breeding success (B)Local availability of food (N)Landscape cover (%)
Crop-types of scenario Current
IBA1.01.05.5
Oat0.50.52.6
Spring barley0.50.50.4
Ryegrass0.20.87.6
Lucerne (Alfalfa)0.20.81.0
Rye0.30.521.3
Triticale0.30.55.3
Winter wheat0.30.58.9
Green rye I0.20.24.4
Green rye II0.20.23.5
Maize (corn)0.20.222.9
Winter canola0.20.28.5
Pasture0.20.28.0
Crop-type of scenario BioenergyPlus
Maize (corn)0.20.245
Green rye I0.20.220
Sudan grass0.20.210
SRC0.00.010
Winter canola0.20.210
Ryegrass0.20.85

The landscape generator's input parameters are average field size, number of crop-types cultivated at the landscape (crop-type richness) and a selection of crop-types that depends on the scenario (see below). Different landscapes, each representing an area of 3 × 3 km, were created in a grid of 750 × 750 patches, resulting in a resolution of 4 × 4 m (e.g. see Fig. 1 and Google, 2010a,b). Landscape creation is a two-step process. First, crop fields of a predefined size are randomly placed consisting of a certain number of neighbouring patches. According to this random positioning, fields may partially overlap which is revised in a subsequent process. This results in an irregular mosaic of fields. The variation of field size around the predefined average is normally distributed. The maximum deviation from predefined field size is around 40–75%, depending on predefined average field size. Second, a crop-type is chosen from a list of up to 13 different crop-types and assigned to each field (see Table 1).

For the analysis of field size and crop richness impact, we used the crop-types of the Current scenario (see section Land-use scenarios and development of nature conservation management strategies and Dziewiaty & Bernardy, 2008). The crop-type assignment process was either random for the analysis of both field size and crop richness impact (see section Impacts of field size and crop-type richness on population density), or created a dedicated area of each crop-type in percent of landscape area (see section Land-use scenarios and conservation options). If the investigated number of crop-types cultivated in the landscape was less than the maximum number of possible crop-types, as with the analysis of reduced crop-type richness, the crop-types used were randomly selected for each individual landscape creation.

Habitat evaluation

A precise definition of regional-scale landscape characteristics is an important basis of habitat evaluation in the light of land-use change analysis. Regarding this issue and to the best of our knowledge, we made a new attempt in defining characteristics which are both related to Skylark population abundance and applicable to the wide range of landscape types we analysed here. We evaluated each virtual landscape based on three landscape characteristics which we deduced from three main requirements of the Skylark. These requirements are associated with both local crop-type characteristics and regional landscape characteristics. For estimation of crop-type parameters ‘probability of breeding successB and ‘local availability of foodN, see section Crop-type parameter estimation, sensitivity analysis and model validation. The three landscape characteristics are the following:

  1. Breeding success (BS): is the probability that a breeding pair successfully reproduces during the current season. The value is determined by the characteristics of the crop-type cultivated at this patch, e.g. crop density, crop height and production system, over the breeding period, taking into account the habitat requirements of the Skylark for a suitable nesting place (Toepfer & Stubbe, 2001) as well as ecological traps (probability of breeding success, B). The latter describes a sudden alteration of the nesting environment resulting in a lower nesting success (Best, 1986; Schlaepfer et al., 2002).
  2. Availability of food (AF): Skylarks seek plants and invertebrates (Holland et al., 2006) within the home range. AF is calculated as the average food availability over all patches within the home range. Food availability is, again, determined by the characteristics of the crop-types cultivated in each patch of the home range (local availability of food, N). The outcome is a value from 0 to 1. Obviously the availability of food could affect the probability of breeding success. However, we suggest this is a minor effect on the probability of successful nests only. Moreover, we indirectly account for this correlation because for some crop-types both values of the crop-type characteristic probability of breeding success (B) and local availability of food (N) are low.
  3. Habitat diversity (HD): Skylarks need food and cover over the entire breeding season, so a high number of different crop-types and various cultivation systems within the home range are favourable (Chamberlain et al., 2000b; Eraud & Boutin, 2002; Vickery et al., 2009). We used the Simpson's diversity index for calculating habitat diversity with respect to the distribution and quantity of crop-types in the patches within the Skylark home range. The outcome is a value from 0 to 1, with 1 being the highest diversity possible. Simpson's index is defined as
    display math
    where pi is the proportion of area of crop-type i within the breeding pair′s home range.

This modelling approach links energy crop cultivation to Skylark habitat suitability. The landscape evaluation was performed in two steps focusing on both habitat suitability and number of Skylark territories.

The first step was the calculation of habitat suitability (HS), based on the three landscape characteristics (BS, AF, HD), for each local patch (750 × 750 patches per landscape) as an index between 0 (habitat is not suitable) and 1 (habitat is very well-suited):

display math

The habitat suitability of the whole landscape was calculated as the average of the habitat suitability of all its component patches. As we used the geometric mean, the habitat suitability was high only when all three landscape characteristics had a high value. The geometric mean is sensitive to low values of a single parameter (here: BS, AF, HD) and is often used in the context of limiting factors in ecosystems (Vinagre et al., 2006; Vincenzia et al., 2006).

For the evaluation of a local landscape patch, we assumed both a nest of a single breeding pair directly on this patch with no other neighbouring breeding pair (thus, potential interference competition is not taken into account, Butler et al., 2010), and a territory of 1.7 ha, which is the average of values found in literature (Paetzold, 1983; Jenny, 1990; Poulsen et al., 1998; Eggers et al., 2011). These simplifications strengthen the focus on habitat evaluation. The evaluation of a certain landscape structure, based on a set of input parameter values including average field size, crop-type number and selected crop-types, was based on the average habitat suitability of 2000 created landscapes.

The second step in habitat evaluation was the explicit spatial distribution of Skylark territories on the landscape with respect to the habitat suitability of each local patch. It is based on the process of immigration to a new habitat. Therefore, the model does not take behaviour like philopatry into account (Greenwood, 1980). This assumption is valid because we aim to simulate the steady state of the abundance after both changes in landscape structure and a related change of skylark population abundance, instead of the transient state.

The date of territory occupation by Skylarks in spring differs for each individual (Delius, 1965). We assumed that best suited habitats are occupied first. Thus, we represented this immigration process by distributing Skylark territories in five steps. First, Skylark home ranges were randomly allocated to patches with a habitat suitability of at least 0.8. The Skylark nest was assumed to be always at the patch in middle of the territory. In the following four steps, Skylark home ranges were distributed to patches with a habitat suitability of at least 0.7/0.6/0.5 and 0.4. As the home ranges of breeding pairs do not overlap with each other, the suitable area for Skylark territories is limited (see Fig. 2 in Jenny, 1990). Thus, the spatially explicit distribution of Skylark breeding pairs was strongly affected by both landscape structure and habitat suitability and provided an approximate estimation of minimum number of breeding pairs in the particular landscape (Fig. 2). For example, due to the home range size of 1.7 ha, the maximum number of breeding pairs was around 5.7 per 10 ha.

Figure 2.

Schematic diagram of the presented methodology for conducting impact assessments. Habitat suitability is calculated based on a virtual landscape (landscape map) and crop-type characteristics (list of crop impacts). The territories of Skylark breeding pairs are placed based on habitat evaluation (habitat suitability), resulting in a spatially explicit distribution and a average Skylark population density for this landscape (abundance).

The analysis of land-use change considered the regional characteristics of different agricultural landscapes. We evaluated this by systematically changing the crop-type richness (the number of crop-types cultivated varies from 2 to 13) and average field size of landscapes (from around 2 ha to 40 ha). We used the number of Skylark breeding pairs as characteristic for landscape evaluation. We also refer to it as Skylark population abundance, however, within a breeding season not every Skylark male could attract a Skylark female, thus our results refer to the reproductive part of the population only (Delius, 1965).

Crop-type parameter estimation, sensitivity analysis and model validation

Crop-type parameter estimation

We suggest the probability of breeding success and local availability of food to be important characteristics of crop-types for the habitat suitability of Skylark. The probability of breeding success and local availability of food of Skylark were not known in detail or poorly understood for the crop-types we used. However, they could be estimated from studies elsewhere, characteristics of the crop species and the usual production system, e.g. sowing and harvesting date. We classified the crop-types into groups and assigned a value from 0 to 1 to each group mostly based on the relative differences between the groups.

Probability of breeding success (B); all crop-types we used were classified in five groups (see Table 1, column two). The classification is based on vegetation density and height (Chamberlain et al., 1999; Donald et al., 2001) within breeding period (Poulsen et al., 1998) and the usual cutting date of crops (Poulsen & Sotherton, 1992; Poulsen et al., 1998). The key characteristics for crop-type classification and the parameter values are summarized in Table 2. Local availability of food (N); all crop-types we used were classified in five groups (see Table 1, column three). The classification is based mainly on pesticide exposure because it impacts Skylark food resources (Aebischer, 1990; Wilson et al., 1999; Vickery et al., 2001; Boatman et al., 2004; Douglas et al., 2010). The key characteristics for crop-type classification and the parameter values are summarized in Table 3.

Table 2. Crop-types grouped by its input parameter value ‘probability of breeding success’. The values were estimated based on literature. Key characteristics of each are presented
Number of crop-type groupCrop-typeEstimated probability of breeding successCharacteristics of this group of crop-typeReferences
B1Integrated Biodiversity Area (IBA)1.0Probability of breeding success is suggested to be very highPoulsen et al. (1998); see ‘habitat pref. index of the Skylark’ Browne et al. (2000)
B2Oat, spring barley0.5Considerably higher territory density in spring-sown compared with autumn-sown cereals throughout the breeding season; vegetation height of spring-sown cereals within breeding period is not very high due to planting in spring (compared with winter cereals) and harvesting is mostly after breeding season ends so we suggest no ecological trapsBest (1986) and USDA (2010)
B3Rye, triticale, winter wheat0.3Rye and triticale are mostly autumn-sown and harvested around the end of breeding period, so we suggest them as an ecological trap. Both cereals grow densely and high within breeding period but seem to provide a little more suitable habitat for Skylark nests than other autumn-sown crop-types we usedSee ‘habitat preference index of the Skylark’ in Browne et al. (2000) and USDA (2010)
B4Ryegrass, Lucerne (Alfalfa), green rye I & II, maize (corn), winter canola, pasture, Sudan grass0.2The crop-types with a probability of breeding success 0.2 are mostly autumn-sown, very dense and tall and are mostly harvested within breeding period (ecological trap) leading to both low nest establishment and low nesting success; within maize fields the predation rate is particularly high after pesticide treatment (which destroys and obstructs vegetation cover; the probability of an unsuccessful nest is particularly high in Lucerne (Alfalfa) and ryegrass fields as they are harvested repeatedly within breeding periodWilson et al. (1997), Dziewiaty & Bernardy (2008), Jenny (1990), Wilson et al. (1997), FSA (2005), USDA (2010)
B5Short-rotation coppice (SRC)0.0SRC (Salix spp.; more than 1 year of growth) is generally avoided by Skylarks as the preferred habitat is farmland without woodSage et al. (2006) and Rowe et al. (2009)
Table 3. Crop-types grouped by its input parameter value ‘local food availability’. The values were estimated based on literature. Key characteristics of each are presented
Number of crop-type groupCrop-typeEstimated local food availabilityCharacteristics of this group of crop-typesReferences
N1Integrated Biodiversity Area (IBA)1.0We suggest no pesticide treatment on IBA thus no food shortage for Skylark; study results show significantly greater arthropod density in IBA than in any other of the cereal fieldsPoulsen et al. (1998)
N2Ryegrass, Lucerne (Alfalfa)0.8Grass species mostly have low pesticide treatments; accessibility of invertebrates to birds may be better in regularly mowed grass than in the tall, dense sward structures of cropsGroom et al. (2008) and Roesch et al. (2009)
N3Oat, spring barley, rye, triticale, winter wheat0.5We suggest these crop-types are mostly cultivated for food production with medium pesticide treatment compared with crop-types of Group 4, however, the level of pesticide treatment may varyGroom et al. (2008)
N4Green rye I & II, maize (corn), winter canola, pasture, Sudan grass0.2Annual crop-types with strong focus on bioenergy production are generally treated with a high input of pesticides; the effects of an intensive management of pasture have been found to generally reduce the number of invertebrate species and, hence, local food availability for the SkylarkGroom et al. (2008), Landis et al. (2008), de Vries et al. (2010), Morris (2000)
N5Short-rotation coppice willow plantation (SRC)0.0Classification of short-rotation coppice willow plantation(with more than one year of growth) is not based on the pesticide exposure as it is generally avoided by Skylarks and hence insufficient for Skylark foragingSage et al. (2006), Rowe et al. (2009, see Table 3)

Sensitivity analysis

We carried out a sensitivity analysis to assess how variation in key parameters affects model predictions of average landscape habitat suitability. Focal parameters were the crop-type characteristics probability of breeding success (B) and local availability of food (N). We calculated the average habitat suitability from 2000 created landscapes with equal input parameters of crop-field size and crop-type richness. The 13 crop-types were distributed both equally, the same area each, and randomly (see section Model structure). To test the sensitivity, we changed the values of the crop-type characteristics probability of breeding success and local availability of food of all crop-types and calculated the habitat suitability of each landscape as mentioned above.

Model validation

We used field data from four different agricultural regions in NE Germany to validate our model results presented in section Impacts of field size and crop-type richness on population density (field data are from the experimental study of Dziewiaty & Bernardy, 2008). The study sites cover areas of 122, 251, 551 and 1058 ha, respectively, which fits well with our focus on virtual landscapes of 900 ha. In each of these regions, Skylark breeding pairs were counted for any crop-type within the 2007 breeding season, and counted at six different fields each, if possible. Crop-types for both bioenergy and food production were considered. We calculated the average Skylark population abundance for each region based on the field size and the number of breeding pairs of all surveyed crop fields. The four study sites differ in both average field size, ranging from 3.7 to 16.9 ha, and number of cultivated crop-types, ranging from 7 to 15. At this level crop number has no considerable impact on Skylark population abundance (note Fig. 4b). We compared two different patterns of field data and model results: Skylark population abundance of a) the landscape (in breeding pairs per 10 ha), and b) the crop fields of a certain size class, in number of breeding pairs per field size class of 2.5 ha range. The latter regards to the fact that crop fields of a certain landscape type, however, differ in size. The analysis of this pattern was based on the data from 90 crop fields of three agricultural regions of the four mentioned above (Dziewiaty & Bernardy, 2008).

Land-use scenarios and development of nature conservation management strategies

Land-use scenarios

We simulated two future scenarios of different land-use changes which would be shaped mostly by potential changes in policy-making and economic aspects of bioenergy production and analysed impacts of the particular landscape structure to Skylark abundance (Boatman et al., 2010). In the Current scenario, we assumed no changes in land use based on present farming practices. For landscape creation we used crop rotation, crop area and land-use data from a region in NE Germany as input data (see Table 1; Dziewiaty & Bernardy, 2008). For the second scenario, BioenergyPlus, we assumed an intensification of bioenergy production including the introduction of some new crop-types compared with the Current scenario, e.g. Sudan grass Sorghum spp. and short-rotation coppice willow plantation Salix spp. (see Table 1 for details of the crop-types we used). Cultivation focuses on fast growing feedstock, so we used six crop-types cultivated for bioenergy production with maize covering a high fraction of the landscape (45%). These assumptions were based on roadmaps and projects of the European Community regarding the creation of ‘bioenergy regions’ in rural areas of Europe with the aim, e.g. to get at least one-third of its energy supply excluding transportation from regional and sustainable biomass sources (SWRA, 2009; BioRegions, 2010). With respect to the variation of regional landscape structures, these scenarios were simulated for two different average field sizes (1.3 and 16.8 ha).

Conservation strategies

For both of the future landscape structure scenarios we analysed two conservation strategies for their ability to compensate biomass production impacts on skylark abundance. Both conservation strategies focus on increasing habitat heterogeneity by patches of near-nature area on bioenergy landscapes (Poulsen et al., 1998; Eraud & Boutin, 2002; Benton et al., 2003; Storkey & Westbury, 2007). We simulated (1) undrilled Patches ‘Skylark plots’ within crop fields of 10% of the landscape area (Odderskaer et al., 1997; Fischer et al., 2009; Schoen, 2011; see Fig. 3c), and (2) IBAs promoted by Cross Compliance Act, a nature conservation act across Europe, with 10% area of agricultural landscape (Henderson et al., 2000).

Figure 3.

Three management options for the application of Integrated Biodiversity Areas (IBA, near-nature area) on a virtual agricultural landscape. The landscape (square) is divided in crop fields of a certain size and shape. Field margins are depicted by black lines and conservation areas by grey lines and squares. Panel (a) IBA on the field margins plus parallel in-field strips of IBA, panel (b) IBA at the field margins plus a medium-sized field used entirely for IBA and panel (c) undrilled patches within crop fields.

Explicit spatial arrangement of conservation areas is suggested to be an important element of conservation success (Vickery et al., 2009; José-María et al., 2010). We analysed the effects of various spatial placements of IBAs to optimize the spatial configuration for particular landscape types. According to the interests of farmers, e.g. easy implementation on fields, IBA could be placed at field margins (Marshall & Moonen, 2002). However, landscapes with large fields do not offer enough area at the margins due to their lower edge-area ratio, e.g. a landscape with average field size of 23 ha provides 3% area for IBA at field margins. So we used additional ‘compensation area’ within crop fields to reach an area of 10%. We implemented: (2a) long thin in-field strips (Fig. 3a, Fischer et al., 2009; Fletcher et al., 2011), (2b) IBA at field margins (Marshall & Moonen, 2002; José-María et al., 2010) and (2c) IBA cultivated on some middle-sized crop fields, replacing the crop-type usually cultivated there (Fig. 3b). As criteria for the ‘best spatial configuration’, we used the Skylark population abundance.

Results

Impacts of field size and crop-type richness on population density

On landscapes with small fields (see Fig. 4a as an example) Skylark abundance declined when crop richness decreased (Fig. 4b), yet with a threshold effect: The decline in abundance was noticeable only when crop richness decreased below a critical value of four crop-types. Above this threshold value, overall abundance was high and mostly insensitive to changes in crop richness. On landscapes with large fields (see Fig. 4c as an example), we observed only a fraction of breeding pairs compared to small-scale landscapes. Here, the impact of crop richness on Skylark abundance was generally low. However, abundance declined with fewer than three crop-types (Fig. 4b).

Figure 4.

General impact of both average field size and number of cultivated crop-types to Skylark population abundance. Panels (a) and (c) show virtual landscapes with different average field size and crop-type richness (see Fig. 1 for more details). Panel (b) presents the abundance of breeding pairs depending on different crop-type numbers (number of crop-types ranges from 2 to 13) and panel (d) presents the abundance of breeding pairs depending on different average field sizes (from 2.4 to 42 ha). Panels (b) and (d) present results for two different types of landscape in detail [see panels (a) and (c) for an example].

For a fixed number of crop-types, an increasing average field size led to a considerable decline in bird abundance (Fig. 4d, results presented for two and for 13 crop-types). Again the results showed a threshold behaviour since an increase of field size in a range from three to around 7 ha had a large impact on Skylark population abundance, whereas an increase in a range from 7 to 40 ha had a much smaller impact.

Our analysis also showed that average field size had a much stronger effect on Skylark abundance, with impact up to two breeding pairs per 10 ha (bp 10 ha−1), than the number of crop-types cultivated with impact up to 1 bp 10 ha−1. Combined effects were found, whereas a fairly high abundance could only be attained if crop-field size was sufficiently small and crop-type richness sufficiently large. For example, Skylark population abundance on a landscape with 9 ha average field size and 13 crop-types (2.6 bp 10 ha−1) would only be slightly lower than on a landscape with a smaller average field size of 3 ha and two crop-type number (2.8 bp 10 ha−1; not shown). So, a decrease of field size would be favourable for Skylark conservation only when number of cultivated crop-types is kept almost constant.

Sensitivity analysis and model validation with field data

The sensitivity analysis considered the effect of two main input parameters of crop-type characteristics on the average landscape habitat suitability. The analysis of probability of breeding success (B) showed a low sensitivity of landscape habitat suitability against this input parameter. The average landscape habitat suitability increased 18% due to the increase of B by 40% and decreased 26% due to the decrease of B by 46%. The second analysis of local availability of food (N), again, showed a low sensitivity of landscape habitat suitability against this input parameter. The average landscape habitat suitability increased 5% due to the increase of N by 15% and decreased 5% due to the decrease of N by 29%. A variation of both crop-type number (from 2 to 13) and average field size (around 2–40 ha) for landscape creation did not significantly changed these results.

We validated the model results using field data of two spatial scales, landscape scale and field scale. Landscape Skylark abundance: Fig. 5a, b shows model results of Skylark population abundance of a landscape (bp 10 ha−1; landscape size 3 × 3 km) and field data of population density on the four different agricultural regions mentioned above. Field data widely supported our model results by matching quite well both the population density of each study site and the decrease of population abundance with increasing average field size. Crop-field Skylark abundance: The average number of Skylark breeding pairs on crop fields classified by size seems to fit well to the observed data, especially for the smaller sized fields. Model results could reproduce the pattern of increase in breeding pair number at larger crop fields (Fig. 6). The noticeable variation in the field data for larger fields might be caused by the low number of data points there.

Figure 5.

Validation of model predictions against empirical data. Average field size effects on Skylark population density (breeding pairs per 10 ha). Panel (a): Field data are illustrated by four closed squares and model results are shown by a dashed (7 crop-types) and a solid line (13 crop-types). Field data are derived from four different regional landscapes of NE Germany (with average field size ranging from 3.7 to 16.9 ha and crop-type number from 7 to 15 crop-types; dataset is from the study of Dziewiaty & Bernardy, 2008). Panel (b): Field data are plotted against model data using four closed squares. Model data are from four virtually created landscapes with the same average field size as the ‘field data landscapes’ and have a crop-type number of 13 each (field dataset is the same as used for panel a). The dashed line (slope is one, crossing both axes at zero) denotes full congruence of model and field data.

Figure 6.

Validation of model predictions for breeding pair number per field size class against field data. Field size effects on number (not density) of Skylark breeding pair per crop fields of a certain size range (arranged in 11 field size ranges). Field data (line with black squares) of Skylark number are sampled on 90 crop fields of three different regions in NE Germany (dataset is from the study of Dziewiaty & Bernardy, 2008).

Land-use scenarios and conservation options

Current and BioenergyPlus scenarios

A considerable increase in energy crop cultivation compared to current practice led to a strong decline in Skylark population abundance (Fig. 7). The relative decline on landscapes with small fields was 82%, and 86% on landscapes with large average field size. The absolute decline on small-scale landscapes is noticeably stronger (−2.4 bp 10 ha−1) than on large-scale landscapes (-1.3 bp 10 ha−1). Thus, the high suitability of small-scale landscapes (see Fig. 4d) is eroded by the increased cultivation of a just few energy crop-types.

Figure 7.

Impact of Current and BioenergyPlus landscape structure and different average field sizes (1.3 and 16.8 ha) on Skylark population density (breeding pairs per 10 ha). Current landscapes reflect the recent crop-type area and crop-type number of a region in NE Germany (13crop-types, see Table 1). BioenergyPlus represents a strong focus on bioenergy production; landscapes are cultivated with six energy crop-types and have a high proportion of maize (45%, see Table 1). Near-nature area (e.g. set-aside) is applied at field margins for Skylark population conservation.

Effects of the conservation options IBA and undrilled patches

Our model results showed that placing IBA on landscapes with small fields leads to small gains for bird abundance on landscapes of the Current scenarios (0.4 bp 10 ha−1) but strongly affects Skylark abundance on BioenergyPlus scenario landscapes (a gain of 2.6 bp 10 ha−1; Fig. 8a). The application of undrilled patches (LW) showed very little change in Skylark abundance (+0.1 bp 10 ha−1) on Current scenario landscapes and a slight increase on BioenergyPlus scenario landscapes (0.8 bp 10 ha−1). However, landscapes with small fields were affected by strong biomass production regardless of the conservation actions. We got a decrease of 0.3 bp 10 ha−1 for IBA and 1.8 bp 10 ha−1 for undrilled patches. The increase of Skylark population density by IBA on Current scenario landscapes with large fields was 0.6 bp 10 ha−1, and 1.9 bp 10 ha−1 on BioenergyPlus scenario landscapes. Our results highlight that strong biomass production has almost no impact when IBAs are applied on landscapes with a large average field size with an impact less than 0.1 bp 10 ha−1 (Fig. 8b). Regardless of the scenario tested, undrilled patches only slightly increased Skylark population density on large-scale landscapes (Current 0.2 bp 10 ha−1; BioenergyPlus 0.5 bp 10 ha−1).

Figure 8.

Impact of Current landscapes (recent landscape structure, 13 crop-types; see Table 1), BioenergyPlus landscapes (cultivation of six energy crop-types, maize has a proportion of 45%) and three different levels of conservation action on Skylark population density (breeding pairs per 10 ha). Levels of conservation action are (a) no conservation action; (b) undrilled patches on selected crop fields and (c) Integrated Biodiversity Area (near-nature area, set-aside) at field margins. Impact is presented for landscapes with a average field size of 1.3 ha (panel a) and 16.8 ha (panel b).

Conservation activities can be spatially arranged in many different ways, often depending on landscape structure and economic needs. The impact of spatially different management strategies including Skylark plots and several options for IBA on landscapes with an average field size of 22.7 ha for Current (light bars) and BioenergyPlus (dark bars) scenario landscapes strongly impacts the Skylark population abundance (see Fig. 9). Skylark abundance was low on landscapes with no conservation activities present, especially in the BioenergyPlus scenario (0.2 and 1.4 bp 10 ha−1 resp., see Fig. 9). The application of both IBAs at field margins and additional in-fieldstrips in order to ensure an IBA-area of 10% (see Fig. 3a) clearly led to a higher Skylark abundance (2.5 and 2.8 bp 10 ha−1 resp.). Furthermore, the effects of a strong increase in energy crop cultivation from Current to BioenergyPlus scenario to Skylark abundance would be reduced from 1.2 to 0.3 bp 10 ha−1, a decrease of 75% relative to ‘no conservation action’ (see Fig. 9). IBA was distributed regionally and, hence, increased the number of Skylark breeding pairs much more than the locally-focused Skylark plots (undrilled patches), particularly at BioenergyPlus landscapes (dark bars of Fig. 9).

Figure 9.

Effect of certain conservation actions on Skylark population density (breeding pairs per 10 ha). Results are presented for two bioenergy production scenarios on a landscape with average field size of 22.7 ha. Dark bars: BioenergyPlus scenario (cultivation of six energy crop-types, maize has a proportion of 45%); light bars: Current scenario (recent landscape structure, 13 crop-types; see Table 1). Upper squares illustrate the distribution of Skylark breeding pairs within the landscape (BioenergyPlus scenario only). Conservation actions (left to right): (a) no conservation action; (b) application of undrilled patches, also referred to as Lark Windows, on certain fields with together 10% area; (c) application of Integrated Biodiversity Areas (IBA, e.g. set-aside) on all field margins (area of 3%); (d) the same as conservation action ‘c’ but additionally complete crop fields cultivated with IBA to get an IBA-area of 10% (see Fig. 3b); and (e) the same as conservation action ‘c’ but additionally strips of IBA within crop fields to get an IBA-area of 10% (see Fig. 3a).

Discussion

In this study, we analysed the effects of energy crop cultivation at the regional landscape scale on the abundance of the ground-breeding farmland bird species Skylark (A. arvensis). Our results showed that the decline of the Skylark population abundance is associated with the land-use change triggered by energy crop cultivation. We suggest that Skylark population abundance is associated with the crop-type richness, the proportion of energy crops and the average field size. Main findings are (a) field size is more important than crop-type richness and (b) the application of IBA on 10% of agricultural landscape would strongly increase Skylark abundance. The results of our study contribute to the debate on the ecological impacts of increased bioenergy production and underline the need of regionally adapted conservation measures.

Impacts of bioenergy production on ground-breeding farmland birds

This class of species is particularly sensitive to changes in agricultural practices and has already shown a strong decline (Ormerod & Watkinson, 2000; Henderson et al., 2009; British Trust for Ornithology, 2011a). These species could therefore be used as indicators of critical ecological trends in agricultural landscapes.

Our model results show that in landscapes with small fields, with an average size of about 5 ha, a decline in crop-type richness at the regional level can substantially diminish the number of breeding pairs, but only if crop-type richness moves below a threshold value of four. However, in accordance with Chamberlain et al. (2000a), Skylark abundance is generally low in landscapes with large fields, with an average size of about 30 ha, irrespective of crop-type richness and the proportion of bioenergy crops. This indicates that the impacts of energy crop cultivation differ, depending on the type of landscape. Considerable impacts of land-use change, e.g. loss of set-aside or decrease of IBA, are more likely in landscapes with small than large fields (see Fig. 8). This highlights the fact that the ecological impacts of biomass production require regionalized assessments that explicitly take the spatial landscape configuration into account (see also Sotherton, 1998; Fischer et al., 2009; Merckx et al., 2009). We therefore suggest that differences of regional landscape types as determined by average field size should be taken into account in bioenergy impacts assessments.

Options for harmonizing biomass production and biodiversity conservation

Our results suggest some options for defining sustainability standards for biomass production. Considering avifauna as an important part of the biosphere (cf. EU Birds Directive) and the strong connection of wild birds to the health of wildlife in the wider countryside (British Trust for Ornithology, 2011a), a certain minimum number of breeding pairs per hectare could be used as a sustainability criterion as in our example of the Skylark: three breeding pairs per 10 ha (Hoffmann, 2007). We demonstrate that the minimum abundance used here is likely to be attained only if both a particular landscape structural diversity and a particular level of crop-type richness can be ensured with focus on the scale of the birds' foraging activity. Evidently, the required minimum abundance translates into a requirement on structural diversity and crop-type richness or sustainability standard.

Our analysis of both conservation measures IBA and undrilled patches show that the application of IBA at field margins are a particularly useful measure for Skylark population conservation. Furthermore, we suggest that negative effects of biomass production can be counteracted by management, even in the BioenergyPlus scenario: (a) in small-scale landscapes, e.g. by adding strips of IBA on field edges; and (b) in large-scaled landscapes, e.g. by combining IBA at field margins with in-field strips of IBA (see Figs 8 and 9). Both undrilled patches and IBA at field margins add habitat diversity and variation in crop structure, which has been suggested as an important option for wild bird conservation (Wilson et al., 2005). However, both facilitate diversity at different spatial scales: undrilled patches locally increase habitat diversity, whereas IBA at field margins form a network that acts regionally. Particularly IBA at field margins has been shown to enhance beneficial species within crops, i.e. biocontrol, thereby reducing pesticide use (Marshall & Moonen, 2002; Landis et al., 2008). A lower pesticide-input, in turn, could reduce the loss of plants and invertebrates by the large-scale energy crop cultivation (Schiesari & Grillitsch, 2011), which is seen as a major factor to farmland bird species abundance decline (Vickery et al., 2001; Boatman et al., 2004).

Potential and limitations of our modelling framework

The spatially explicit approach of our model allows the regionalized and spatially differentiated analysis of both the impact of agricultural land-use change triggered by large-scale energy crop cultivation and the effectiveness of conservation management measures over one breeding season of Skylark. The modelling approach can be applied to both real and hypothetical virtual landscapes. The possibility of analysing systematically varied, virtual landscapes is one of the strengths of the presented approach as it provides insight into what situations are critical for the conservation of farmland bird species population abundance and why. The creation of virtual agricultural landscapes and the spatially explicit assessment helps to overcome limitations of existing models which are nonspatial or focus on a specific area (Butler et al., 2007; Firbank, 2008; Boatman et al., 2010; Zhang et al., 2010). A special feature of the present framework are the different sources of diversity. In this case, landscape structural diversity and crop-type richness can be assessed in terms of their importance for enabling bird species to sustain a viable abundance. Our modelling framework contributes to questions addressing ecological aspects of bioenergy feedstock production at the regional scale of landscape such as ‘How much near-nature area is required to sustain a Skylark population in a certain type of landscape?’ or ‘Where should a strong increase of energy crop cultivation be excluded and where should it be promoted?’

One common approach for assessing the impacts of landscape structural changes on a certain species is habitat modelling. Within this approach, knowledge of the demands of the species and statistical analyses are combined to predict suitable habitat and species occurrence in relation to environmental conditions in a spatially differentiated way (Schadt et al., 2002; Elith et al., 2006). A weakness of this approach is its static character. In an agricultural landscape, environmental conditions can change rapidly from year to year as result of both cultivation of a certain crop species and the crop production system applied. This can strongly affect all demographic characteristics of the individual bird species such as survivorship, i.e. probability of ecological traps, or food availability such as change in food supply in the surroundings. These affect the breeding success of the breeding pair and so the bird population abundance. We take a step forward by developing this model framework with habitat modelling as a basis and incorporating some features of dynamic animal population modelling.

However, the focus on one breeding season keeps the model simple. Moreover, the current version of the model is simplified by the assumption of a fixed territory size. Territory size is often variable and adapted to food supply, so is has to be tested to what extent such modifications would alter the results (Butler et al., 2010). Though, we have shown that in the case of Skylark (A. arvensis), there is a good correspondence between simulated and measured average abundance, despite the simplicity of the model (see Fig. 5).

The model could easily be extended to a set of species or functional species. This would allow the assessment of the impacts of biomass production on a larger number of different farmland bird species and could provide answers to questions related to sustaining farmland biodiversity.

Acknowledgements

The study was supported by the Helmholtz Research Programme ‘Terrestrial Environments: Strategies of Response to Global and Climate Change’ and the Federal Agency for Nature Conservation (Project code: 3507 82 150). We are particularly grateful to K. Schümann for ideas and suggestions formulated during the development of this study and to J. Calabrese for editing the manuscript. We are also very grateful to the anonymous referees for valuable comments that significantly improved the manuscript.

Ancillary